1

On multiple wh-fronting*

�eljko Bo� kovi �

It is standardly assumed that there are 4 language types with respect to possibiliti es for wh-

movement in multiple questions. English-type, where only one wh-phrase moves, Chinese-type,

where they all stay in situ,1 and French-type, where both of these options are available, are

illustrated in (1)-(3).

(1) What did John give to who?

(2) John gei-le shei shenme?

John give perf. who what

‘What did John give to who?’

(3) a. Qu’ a-t-il donné à qui?

what has he given to who

b. Il a donné quoi à qui?

he has given what to who

French is often assumed to be a simple mixture of the first two types. This view is mistaken. If it

were correct both the English and the Chinese strategy would always be possible in French, i.e. the

set of possibiliti es for questions in French would be a union of the set of possibiliti es for questions

in English and Chinese. This is not true. As shown in Bo� kovi � (1998a, 2000), the in-situ strategy

has a very limited distribution in French, which indicates that its wh-in-situ is of different nature

from Chinese wh-in-situ. Based on this, I assume that French is a separate type, not a simple mixture

of English and Chinese. The relevant French data are given in (4). Wh-in-situ is allowed in short

distance null C matrix questions (3b), but not in embedded, long-distance, and overt C questions (4).

(See Bo� kovi � 1998a, 2000 for explanation for the limited distribution of wh-in-situ in French. The

judgments are given for the true question reading. Note that overt C questions are possible only in

some dialects.)

(4) a. *Pierre a demandé tu as embrassé qui.

Peter has asked you have kissed who

2

b. cf. Pierre a demandé qui tu as embrassé.

c. *Jean et Marie croient que Pierre a embrassé qui?

John and Mary believe that Peter has kissed who

d. cf. Qui Jean et Marie croient-ils que Pierre a embrassé?

e. * Que tu as vu qui?

C you have seen who

f. Qui que tu as vu?

The fourth type, multiple wh-fronting (MWF) languages, where all wh-phrases move, is ill ustrated

by (5) from Bulgarian, where according to Rudin (1988) all wh-phrases move to SpecCP overtly.

(5) Na kogo kakvo dade Ivan?

to who what gave Ivan

‘What did Ivan give to who?’

This paper deals with MWF. In section 1 I argue that the MWF-type should be eliminated from the

above typology, languages considered to belong to this type being scattered across the first three

types. In section 2 I show that there are several classes of non-MWF questions in MWF languages,

a surprising fact given the discussion in section 1, which will l ead me to posit a new type of in situ

wh-phrases not attested in English, French, and Chinese-type languages. Section 3 is the conclusion.

1 When MWF languages have wh-movement

1.1 Superiority effects in MWF languages

One argument that MWF languages are scattered across the English, French, and Chinese-type with

respect to when they have wh-movement concerns Superiority effects, reflected in the order of

fronted wh-phrases.2 There are three types of MWF languages with respect to Superiority. Serbo-

Croatian (SC) exhibits Superiority effects in some contexts. Bulgarian exhibits them in all contexts,

while Russian never exhibits them. Consider first SC. SC exhibits Superiority effects in embedded,

long-distance, and overt C questions, but not in short-distance null C matrix questions:3

(6) a. Ko koga voli?

3

who whom loves

‘Who loves whom?’

b. Koga ko voli?

(7) a. [Ko koga voli], taj o njemu i govori.

who whom loves that-one about him even talks

‘Everyone talks about the person they love.’

b. ?*[Koga ko voli], taj o njemu/o njemu taj i govori.

(8) a. ?Ko koga ka� e� da je istukao?

who whom say that is beaten

‘Who do you say beat whom?’

b. *Koga ko ka� e� da je istukao?

(9) a. (?)Ima ko � ta da ti proda.

has who what part. you sells

‘There is someone who can sell you something.’

b. *Ima � ta ko da ti proda.

(10) a. Ko li koga voli?

who C whom loves

‘Who on earth loves whom?’

b. *Koga li ko voli?

Notice that SC exhibits Superiority effects exactly where French must have wh-movement. Where

French does not have to have wh-movement, SC does not exhibit Superiority effects.

Bulgarian exhibits Superiority effects in all contexts, including (6)-(10) (see also (41b)/(43b)).

(11) a. Koj kogo obi� a?

who whom loves

b.*Kogo koj obi� a?

c. Koj kogoto obi� a, toj za nego i govori.

who whom loves he about him even talks

d. *Kogoto koj obi� a, toj za nego/za nego toj i govori.

4

e. Koj kogo kazva� � e e nabil?

who whom say that is beaten

f. *Kogo koj kazva� � e e nabil?

g. (?)Ima koj kakvo da ti prodade.

has who what part. you sells

h. *Ima kakvo koj da ti prodade.

i. Koj li kogo obi� a?

who C whom loves

j. *Kogo li koj obi� a?

Finally, as shown by Stepanov (1998), Russian has free order of fronted wh-phrases in all contexts.

This is ill ustrated in (12) for the contexts in (6)-(9). (Notice that Russian does not allow wh-phrases

in the li -construction and does not allow multiple questions in the wh-existential construction.)

(12) a. Kto kogo ljubit?

Who whom loves

b. Kogo kto ljubit?

c. Kto kogo uznaet, tot togo i poljubit.

who whom knows that-one.nom that-one.acc and loves

‘Everyone will love the person they will know.’

d. Kogo kto uznaet, togo tot i poljubit.

e. Kto kogo ty xo� e� , � toby pobil?

who whom you want that-subj. beat

‘Who do you want to beat whom?’

f. Kogo kto ty xo� e� , � toby pobil?

There is a parallelism in the behavior of English, French, and Chinese with respect to wh-movement

and the MWF languages with respect to Superiority: SC exhibits Superiority effects where French

must have wh-movement, Bulgarian where English must have wh-movement (all contexts), and

Russian where Chinese must have wh-movement (namely, never).4 This can be accounted for if SC,

5

Bulgarian, and Russian behave like French, English, and Chinese with respect to when they must

have wh-movement, which I take to be movement motivated by checking the strong +wh-feature of

C: SC must have it in long-distance, embedded, and overt C questions, but not in short-distance null

C matrix questions; Bulgarian must have it in all contexts, and Russian does not have to have it at

all (see also Stepanov 1998 and Strahov 2000 for Russian.) Wh-movement in MWF languages is

then well -behaved with respect to Superiority: SC, Bulgarian, and Russian exhibit Superiority effects

whenever they have wh-movement. The only difference between SC/Bulgarian/Russian and French/

English/Chinese is that even wh-phrases that do not undergo wh-movement in the former group still

must be fronted for independent reasons. That this movement is not driven by the strong +wh-feature

of C is confirmed by the fact that all wh-phrases must move in these languages, although movement

of one wh-phrase should suff ice to check the strong +wh-feature of C. (I refer to this obligatory

movement of wh-phrases that is independent of the strong +wh-feature of C as non-wh-fronting.)

(13) a.Ko ta kupuje? (SC)

who what buys

'Who buys what?'

b. ?*Ko kupuje ta?

(14) a. *Koj e kupil kakvo? (Bulgarian)

who is bought what

‘Who bought what?’

b. Koj kakvo e kupil?

(15) a. *Kto kupil to? (Russian)

who bought what

b. Kto to kupil?

Even echo wh-phrases must move in these languages. (13b), (14a), (15a), and (16) are unacceptable

even as echo questions (see also Wachowicz 1974 and Kiss 1987 for Polish and Hungarian), which

confirms that wh-phrases in MWF languages front independently of the strong +wh feature of C.5

(16) a. ?*Ivan kupuje ta? (SC)

6

Ivan buys what

b. ?*Ivan e kupil kakvo? (Bulgarian)

Ivan is bought what

c. ?*Ivan kupil � to? (Russian)

Ivan bought what

Stjepanovi� (1998, 1999b) argues the driving force of SC non-wh-fronting is focus, SC wh-phrases

being inherently focused. She follows the line of work originating with Horvath (1986), where wh-

fronting in a number of languages is analyzed as focus movement. This work makes a correlation

between movement of wh-phrases and movement of contrastively focused non-wh-phrases,6 whereby

a number of languages that overtly move non-wh-phrases with this type of focus are analyzed as

having focus fronting of wh-phrases. (I will refer to contrastively focused non-wh-phrases simply

as focused, this being the only type of focus for such phrases I am concerned with.)The analysis has

been convincingly applied to, e.g., Aghem, Basque, Hungarian, and Quechua (see, e.g., Horvath

1986, Rochemont 1986, and Kiss 1995.) Stjepanovi � shows that SC fits into this line of research.

We have already seen that SC fronts all wh-phrases. It also fronts focused non-wh-phrases, given

in capitals.7

(17) a. JOVANA savjetuje.

Jovan.acc advises

‘(S)he advises Jovan.’

b. ?*Savjetuje JOVANA.

Stjepanovi� provides convincing evidence that focused non-wh-phrases and wh-phrases undergo the

same kind of movement in SC based on adverb placement. The focus movement analysis is applied

to Bulgarian in Bo kovi � (1998b, 1999), Izvorski (1993), and Lambova (in press), Russian in

Stepanov (1998), and Romanian, also a MWF language, in Göbel (1998).8 I will also adopt it here.

Notice, however, that my conclusions concerning when MWF languages have wh-movement are

unaffected by the precise identity of the driving force of non-wh-fronting. However, I provide below

three additional arguments for the focus movement analysis concerning D-linked and echo wh-

7

phrases (section 2.1) and the distribution of parentheticals in questions (footnote 15). Before doing

that, in the next section I present evidence concerning the interpretation of multiple questions which

confirms the conclusion reached above based on Superiority with respect to when various MWF

languages must have wh-movement. (Another argument concerning Superiority is given in section

2.2.)

1.2 Interpretation of multiple questions

It is well -known that a pair-li st answer is obligatory for constructions like (18). (The observation is

due to Wachowicz 1974. See also Comorovski 1996:44, who explains away an exception to the

observation concerning reversible predicates.) Thus, (18) cannot be felicitously asked in the

following situation: John is in a store and sees somebody buying an article of clothing, but does not

see who it is and does not see exactly what the person is buying. He goes to the sales clerk and asks

(18).

(18) Who bought what?

Interestingly, single-pair answers are not crosslinguistically infelicitous with questions like (18).

Thus, Japanese (19) can have either a single-pair or a pair-li st answer, as observed by Mamoru Saito

(p.c.). The example can be used in the situation described above, in contrast to (18).

(19) Dare-ga nani-o katta no?

who-nom what-acc bought Q

‘Who bought what?’

Chinese and Hindi pattern with Japanese. German, on the other hand, patterns with English. An

obvious difference between English/German and Japanese/Chinese/Hindi is that the former have

overt wh-movement, whereas the latter are wh-in-situ languages; i.e., interrogative SpecCPs must

be fill ed overtly by a wh-phrase in English and German, but not in Japanese, Chinese, and Hindi.

(I ignore the possibilit y of null operator movement, focusing on wh-phrases.) It is possible that overt

movement to SpecCP forces pair-li st answers. French, which can employ either the in-situ or the wh-

movement strategy, confirms the conjecture. Single-pair answers are possible in French, but only

8

with in-situ questions. Thus, the in-situ question in (20a) can have a single-pair answer, which is not

possible with (20b). (I discuss only non-subject questions in French, where it is clear when wh-

movement occurs.)

(20) a. Il a donné quoi à qui?

he has given what to who

‘What did he give to who?’

b. Qu’a-t-il donné à qui?

The contrast between (20a) and (20b) strongly indicates that the availabilit y of single-pair answers

depends on the possibility of not moving any wh-phrase to SpecCP overtly.9

Turning to Slavic, as expected, Bulgarian, a language in which interrogative SpecCPs are

obligatorily fill ed by a wh-phrase overtly, patterns with English in that (21) requires a pair-li st

answer.

(21) Koj kakvo e kupil?

who what is bought

‘Who bought what?’

Significantly, SC patterns with languages in which wh-phrases do not have to move to SpecCP

overtly. Thus, SC (22) can have either a pair-li st or a single-pair answer. This indicates that SC

questions are well-formed even when no wh-phrase moves to interrogative SpecCP overtly.10

(22) Ko je � ta kupio?

who is what bought

‘Who bought what?’

Stepanov (1998) notes that Russian questions like (23) also allow single-pair answers, as expected.

(23) Kto � to kupil?

who what bought

‘Who bought what?’

9

Polish and Romanian confirm the analysis. Like SC and Russian, Polish does not show Superiority

effects in short-distance null C questions (see Rudin 1988), which means that it does not have to

have overt wh-movement in such questions. On the other hand, Romanian shows Superiority effects

(see Rudin 1988), which means that it has obligatory overt wh-movement, li ke Bulgarian and

English.

(24) a. Kto co kupi� ? (Polish)

who what bought

‘Who bought what?’

b. Co kto kupi� ?

(25) a. Cine ce a cump� rat? (Romanian)

who what has bought

‘Who bought what?’

b. *Ce cine a cump� rat?

Significantly, Citko and Grohmann (2000) observe that a single pair answer is possible with Polish

(24a), but not with Romanian (25a), a strong confirmation of the current analysis (see Bo� kovi � in

press and Citko and Grohman 2000 for discussion of the interpretation of (24b).)

2 In situ wh-phrases in MWF languages

In this section I show that there are some exceptions to the obligatoriness of fronting of wh-phrases

in MWF languages, a surprising fact in light of the above discussion. The exceptions can be

classified into three groups: semantic, phonological, and syntactic. I start by examining semantic

exceptions.

2.1 Semantic exceptions to the obligatoriness of wh-fronting in MWF languages

One semantic exception involves D-linked wh-phrases, which can remain in situ, as shown in (26).11

(26) a. Ko je kupio koju knjigu? (SC)

who is bought which book

10

‘Who bought which book?’

b. Koj e kupil koja kniga? (Bulgarian)

who is bought which book

c. (?)Kakoj student pro� ital kakuju knigu? (Russian)

which student read which book

The exceptional behavior of D-linked wh-phrases is explainable under the focus analysis. With D-

linked wh-phrases the range of felicitous answers is limited by a set of objects familiar to the speaker

and the hearer as a result of it being referred to in the discourse or salient in the context. The range

of reference of D-linked wh-phrases is thus discourse given. Due to their "discourse giveness", such

wh-phrases are not inherently focused hence should not be subject to focus movement.12 Note that

some speakers prefer leaving D-linked wh-phrases in situ.Wachowicz (1974) notes this for Polish

and Pesetsky (1987, 1989) for Romanian and Russian.Some speakers, however, can optionally front

them. Thus, SC (27) is only slightly degraded.(Some Polish, Russian, and Romanian speakers allow

(27).)13

(27) ?Ko je koju knjigu kupio?

who is which book bought

It is plausible that the D-linked wh-phrase in (27) undergoes scrambling rather than focus movement.

If the latter were the case we would expect the movement to be obligatory, which is not the case with

scrambling, scrambling being optional.14 This means that at least marginall y, wh-phrases can be

scrambled in SC (see Sinicyn 1982 for Russian). Notice that there is crosslinguistic variation in this

respect. Thus, Japanese allows wh-scrambling, while German does not (see Müller and Sternefeld

1996). The scrambling analysis thus may make it possible to account for the variation regarding (27).

An interesting confirmation of this analysis is provided by Bulgarian, where most speakers

allow optional fronting of D-linked wh-phrases under consideration (cf. (26b)).

(28) Koj koja kniga e kupil?

who which book is bought

11

Rudin (1988) argues that in Bulgarian constructions like (29), all wh-phrases are located in SpecCP,

which in our terms means that the focus licensor for Bulgarian wh-phrases is interrogative C. One

argument for Rudin’s claim concerns the fact that the wh-phrases cannot be split by a parenthetical.15

(29) Koj kakvo e kupil?

who what is bought

(30) ?*Koj, spored tebe, kakvo e kupil?

who according you what is bought

‘Who, according to you, bought what?’

Significantly, it is easier to split wh-phrases with a parenthetical i f the second wh-phrase is D-linked.

(31) ?Koj, spored tebe, koja kniga e kupil?

who according you which book is bought

‘Who, according to you, bought which book?’

(30)-(31) provide evidence that kakvo in (29) and koja kniga in (28) are not located in the same

position, which follows if, in contrast to kakvo in (29), koja kniga in (28) does not undergo focus

movement, remaining below CP. I conclude, therefore, that D-linked wh-phrases do not undergo

non-wh-fronting, which is explained under the focus movement analysis of non-wh-fronting.

Notice now that if, as is often assumed, Engli sh does covertly what Slavic languages do

overtly with respect to wh-phrases, only non-D-linked wh-phrases would have to undergo LF

movement in English, as argued in Pesetsky (1987) (see also Bo� kovi � and Franks 2000). However,

they would undergo focus movement, not wh-movement.

A question arises whether a D-linked wh-phrase can stay in situ in a single question. This is

not completely clear in SC. (32) is degraded on the true question reading, but not fully unacceptable.

(32) ??On je kupio koju knjigu?

he is bought which book

‘He bought which book?’

12

I suggest that the degraded status of (32) on the true question reading is a result of the failure to type

the clause as a question in the sense of Cheng (1997), who argues that each clause must be typed,

i.e., identified as declarative or interrogative, in overt syntax.16 Interrogative identification is done

either through a question particle or by fronting a wh-phrase. Since SC does not have a question

particle in the relevant constructions, one wh-phrase must front for typing purposes. I assume the

typing is carried out by fronting and pronouncing a wh-phrase within the highest phonologicall y

realized projection in overt syntax.17 In D-linking questions this can be done through either

scrambling or wh-movement. (Given that SC patterns with French with respect to when it has overt

wh-movement, wh-movement should be an option even in short-distance matrix questions like (6a)

since in French wh-movement takes place optionally in such questions. Recall that overt wh-

movement cannot have taken place in the grammatical derivation of (6b) because of Superiority.)

I assume that when wh-phrases in questions like (6) (or Russian (12)) are not D-linked the typing

can be carried out within the focus-licensing projection, which can be the highest projection given

that, as argued in Bo� kovi � (1997c, 2000), CP does not have to be inserted until LF in (6). I argue

that +wh C in questions like (6a) can be inserted either overtly or covertly. If inserted covertly, no

overt wh-movement takes place. If inserted overtly, wh-movement takes place overtly.(I show that

in structures where the wh-movement option is forced LF C-insertion is blocked.)18 One argument

for the analysis not noted in the works cited above concerns topic constituents(TC)(see Stjepanovi �1999a,b for another argument based on sluicing). With TCs, SC shows Superiority effects even in

short distance null C questions.

(33) a. Tom � oveku, ko je � ta poklonio?

that man.dat who is what bestowed

‘To that man, who bestowed what?’

b. ??Tom � oveku, � ta je ko poklonio?

Rudin (1993) discusses TCs in Bulgarian and argues that TCs are adjoined to CP.19 TCs can then be

present in the structure only when CP is present overtly. Overt insertion of a +wh C induces a

Superiority effect, which means that it forces wh-movement. It follows then that in (6a), which does

not show Superiority effects hence does not involve overt wh-movement, CP is not inserted overtly.

13

Notice that Russian does not exhibit Superiority effects even in TC constructions. This is

expected given that Russian is a Chinese-type language with respect to when it must have wh-

movement. Russian never has wh-movement regardless of whether C is inserted overtly or covertly.

(34) a. A etomu � eloveku kto kogo predstavil?

and that man.dat who whom introduced

‘And to that man, who introduced whom?’

b. A etomu � eloveku kogo kto predstavil?

Notice also that, in contrast to (22), SC (33a) can only have a pair-li st answer. This is expected.

Recall that TCs require overt C-insertion, which in turn triggers overt wh-movement. In contrast to

(22), (33a) then must involve overt movement to SpecCP, hence the obligatoriness of a pair-li st

answer. On the other hand, Stepanov (1998) notes that Russian (34a) can still have a single-pair

answer, as expected given that Russian questions do not have to involve overt movement to SpecCP.

Returning to the typing requirement, notice that although (35a) is bad on the true question

reading, it is good on the echo reading. (The judgment holds for the request for repetition reading.)

(35) a. Ona tvrdi da ŠTA/?*šta mrzi?

she claims that what hates

‘She claims that she hates WHAT?’ /‘What does she claim that she hates?’

b. ?*Ona tvrdi da mrzi ŠTA/šta?

This is expected. The current analysis attributes the badness of (35a) on the true question reading to

the failure to type the matrix clause as interrogative. Since echo questions are not subject to the

typing requirement (note that English echo questions do not have to involve overt movement), the

problem does not arise on the echo reading. The wh-phrase still has to front for the reason discussed

above (focus), which does not apply in English and which is independent of the typing requirement

in the sense that wh-phrases are subject to it in MWF languages even when the typing is not an issue.

Notice that echo questions like (36) and (35b) are significantly better (in fact acceptable) on

the reading on which they express surprise than on the reading on which they ask for repetition of

14

what the echo questioner has not heard. (The judgment is given for the latter reading. The surprise

reading generally induces even stronger stress on the echo wh-phrase than the request for repetition

reading. For discussion of different types of echo questions, see Pope 1976 and Wachowicz 1974.)

(36) ?*Ona je poljubila KOGA?

she is kissed who

‘She kissed WHO?’

This can be straightforwardly accounted for under the focus movement analysis since the value of

the echo wh-phrase is fully known to the speaker, as well as the hearer, on the surprise reading, but

not on the request for repetition reading. (Recall that focus represents new information.)

To sum up, in contrast to non-D-linked wh-phrases, D-linked wh-phrases and certain echo

wh-phrases can remain in situ in MWF languages, which can be accounted for under the focus

movement analysis, a fact that should be interpreted as evidence for the analysis. (Recall that the

possibilit y of focused material splitti ng fronted non-D-linked wh-phrases in Bulgarian also provides

evidence for the focus movement analysis.) We have seen that there are three distinct ways of

fronting wh-phrases in MWF languages: wh-movement, pure focus movement, and scrambling. The

second way is the only one that is always fully acceptable for all speakers of MWF languages.20

2.2 Phonological exceptions to the obligatoriness of wh-fronting in MWF languages

I turn now to phonological exceptions to the obligatoriness of fronting wh-phrases in MWF

languages. One such exception is exempli fied by SC (37), which contrasts with (13b) and (16a).21

(37) � ta uslovljava � ta?

what conditions what

What is at stake here is the actual phonological form of the wh-phrases. The second wh-phrase does

not front if it is homophonous with the first fronted wh-phrase.22 Apparently, SC does not allow

sequences of homophonous wh-words. To avoid formation of such a sequence a wh-phrase can

remain in situ. Notice that in (38) the second wh-phrase must front. Because of the adverb, fronting

15

of the second � ta does not create a sequence of homophonous wh-words.

(38) a. � ta neprestano � ta uslovljava?

what constantly what conditions

‘What constantly conditions what?’

b. ?*� ta neprestano uslovljava � ta?

Leaving a wh-phrase in situ thus can be done only as a last resort when necessary to avoid forming

a sequence of homophonous wh-words. The same holds for Bulgarian, Russian, and Romanian.

(39) a. Kakvo obuslavlja kakvo? (Bulgarian)

what conditions what

b. *Kakvo kakvo obuslavlja?

c. to obuslovilo ! to? (Russian)

what conditioned what

d. * to ! to obuslovilo?

e. Ce precede ce? (Romanian)

what precedes what

f. *Ce ce precede?

We seem to be dealing here with a low level PF effect, since the information concerning the

pronunciation of wh-phrases should not be accessible to the syntax. It appears that we need a PF

constraint against consecutive homophonous wh-phrases in the languages in question. Billi ngs and

Rudin (1996) in fact propose such a constraint for Bulgarian to account for (40a):23

(40) a. *Koj na kogo kogo e pokazal?

who to whom whom is pointed-out

‘Who point out whom to whom?’

b. Cf. Koj kogo na kogo e pokazal?

Notice that we cannot be dealing here with a Superiority effect. Bo" kovi # (1997b) shows that only

16

the highest wh-phrase is sensitive to Superiority in Bulgarian, i.e. the highest wh-phrase moves first,

the order of movement of other wh-phrases being in principle free. This is shown by (41)-(44).

(Recall that the order of wh-phrases corresponds to the order of movement to SpecCP.)24

(41) a. Kogo kak e tselunal Ivan?

whom how is kissed Ivan

‘How did Ivan kiss whom?’

b. ?*Kak kogo e tselunal Ivan?

(42) a. Koj kogo kak e tselunal?

who whom how is kissed

‘Who kissed whom how?’

b. Koj kak kogo e tselunal?

(43) a. Kogo kakvo e pital Ivan?

whom what is asked Ivan

‘Whom did Ivan ask what?’

b. ?*Kakvo kogo e pital Ivan?

(44) a. Koj kogo kakvo e pital?

who whom what is asked

‘Who asked whom what?’

b. Koj kakvo kogo e pital?

Note now the parallelism between the wh-phrases in SC (6) and non-initial fronted wh-phrases in

Bulgarian with respect to the lack of Superiority effects. The parallelism is expected under the

current analysis, where movement of the first wh-phrase in Bulgarian differs from the movement of

the second and the third wh-phrase, which are in turn the same as the movement of all wh-phrases

in SC (6). In other words, since the second and third movement in Bulgarian (41)-(44) and both

movements in SC (6) are the same (both can be pure focus movement) it is not surprising that they

behave in the same way with respect to Superiority, differing in this respect from the first movement

in (41)-(44). The data show that only the wh-phrase checking the strong +wh-feature of C (which

means only one wh-phrase) is subject to Superiority, wh-phrases undergoing pure focus movement

17

being insensitive to it. Bo$ kovi % (1999) gives an economy-based explanation of this fact applicable

to both SC and Bulgarian that is based on certain differences in formal properties of focus and wh-

movement.25

Interestingly, li ke Bulgarian, SC exhibits selective superiority effects where it must have wh-

movement, as in, e.g., embedded wh-clauses. As in Bulgarian questions, in such contexts the highest

wh-phrase prior to movement is first in the linear order, the order of other wh-phrases being free.26

(45) a. ?Ima kome kako da pomogne.

has whom how part. helps

‘S(he) has someone to help somehow.’

b. *Ima kako kome da pomogne.

c. ?Ima ko kako kome da pomogne.

has who how whom part. helps

‘There is someone who can somehow help somebody.’

These facts confirm the current analysis of when SC and Bulgarian must have overt wh-movement.

Returning to (40), given the above discussion, (40) cannot be accounted for by Superiority.

Notice also that (40a) improves when the third wh-phrase remains in situ, which is not unexpected

if the unacceptability of (40a) is indeed due to a PF constraint against homophonous sequences of

wh-phrases. The third wh-phrase cannot remain in situ in (40b), which confirms that leaving a wh-

phrase in situ is a last resort device for saving a question from violating the PF constraint in question.

(46) a. Koj na kogo e pokazal kogo?

who to whom is shown whom

‘Who showed whom to whom?’

b. ??Koj kogo e pokazal na kogo?

We are dealing here with a rather intricate interplay of phonology (the PF constraint in question) and

syntax (the obligatoriness of fronting wh-phrases). A principled way of capturing it is provided by

a recent approach to the pronunciation of non-trivial chains, based on the copy theory of movement.

18

It is generally assumed that on the LF side, we have some choice in deciding where deletion should

take place in non-trivial chains. Thus, Chomsky (1995) argues that on the reading on which himself

in (47) refers to Joe, the tail of the chain created by wh-movement of which picture of himself is

deleted so that himself remains in SpecCP, where it is c-commanded by Joe but not Jim. On the other

hand, on the reading on which himself refers to Jim, himself is deleted in the head of the chain and

remains in the structure in the tail of the chain, where it is c-commanded by and local to Jim.27

(47) Joe wonders [CP [which picture of himself] [IP Jim bought [which picture of himself]]]

In LF we thus have a choice in deciding where deletion should take place in non-trivial chains. It is

often assumed that no choice is available in PF, the head always being the sole survivor, as in (48).

(48) a. The student was arrested the student.

b. *The student was arrested the student.

c. *The student was arrested the student.

d. *The student was arrested the student.

However, a number of authors have recently argued that in PF we also have a choice concerning

which member of a non-trivial chain survives deletion (see Bobalji k 1995, Brody 1995, Groat and

O’Neil 1996, Runner 1998, Hiramatsu 1997, Pesetsky 1997,1998 Richards 1997, Roberts 1997,

Franks 1998, Nunes 1999, and Bo& kovi ' 2001). Of particular interest to us is the proposal made in

Franks (1998), who argues that, just as in LF there is a preference for deletion in the head position

of non-trivial chains (at least with operator-variable chains), in PF deletion in the tail of non-trivial

chains (that is, deletion of lower copies) is just a preference. More precisely, a lower member of a

chain is pronounced instead of the head of the chain iff pronunciation in the head position would lead

to a PF violation, provided that the violation can be avoided by pronouncing the lower member of

the chain.28 (By the head of a chain I mean the highest member of a sequence of copies created by

movement of the same element. I disregard the fact that in some cases two different chains, an A and

an A’-chain, are created by movement of the same element, as in Whoi ti seems ti to ti know it.)

Let us see what the proposal entails for (37). The ungrammaticality of (13b) and (16a) shows

that there is a syntactic requirement, namely focus, that forces all wh-phrases in SC to move in overt

19

syntax. This should also hold for the second ( ta in (37), which then also must undergo focus

movement. As a result, (37) has the following SS. (I am ignoring the lower copy of the first ( ta.)

(49) [FP ) ta * tai [uslovljava * tai]]

what what conditions what

I assume that there is a PF constraint against consecutive homophonous wh-words in SC. Given the

constraint and the proposal that a lower member of a non-trivial chain can be pronounced if

necessary to avoid a PF violation, we are allowed to pronounce the lower copy of the second ( ta in

PF.29

(50) [FP ) ta * tai [uslovljava * tai]]

We thus derive (37) and account for the contrast between (37) and (13b)/(16a) without violating the

syntactic requirement that forces all wh-phrases to move overtly in SC (the second ( ta in (37) does

undergoes focus movement), without look-ahead from the syntax to the phonology, and without any

PF movement. The analysis also provides evidence for the copy theory of movement.

Consider now Bulgarian (46). (46a-b) have the SS in (51). (The order of the objects in the

base position and the precise position of the subject prior to wh-movement are irrelevant. Recall that

the order of wh-phrases reflects the order of movement to SpecCP. Koj moves first, the order of

movement of the objects is free. Pokazal undergoes short V-movement and e may be moving to C.)

(51) a. Koji na kogoj kogok e koji pokazal na kogoj kogok?

b. Koji kogoj na kogok e koji pokazal na kogok kogoj?

Consider which copies of the wh-chains will be pronounced in (51a). Since we are dealing with a

PF operation, it seems natural to scan the structure linearly left-to-right. We then first examine the

koj chain. Since nothing goes wrong if the chain is pronounced in the head position, we pronounce

the initial koj. Next, consider the na kogo chain. Again, no PF violation occurs if we pronounce its

head. (Nothing rules out the koj na kogo sequence. Note that I assume that the decision whether to

pronounce the head or the tail of the chain is made without look-ahead. It cannot be affected by later

decisions concerning pronunciation of other chains.)At this point, then, we have the sequence koj

20

na kogo sentence initially. Now we consider the kogo chain: If we pronounce kogo in the head

position we violate the PF constraint against homophonous sequences of wh-words. In order not to

do that we pronounce the tail of the chain, deriving (46a). Consider now (51b). It is easy to verify

that if we scan the structure left to right when determining which copies to pronounce, no PF

violation occurs if we pronounce the heads of all three chains. We then must pronounce the initial

wh-phrases, which gives us (40b). Note that (46b) is underivable. The data in (40) and (46) are thus

accounted for.

Romanian, a Bulgarian-type MWF language (see Rudin 1988), provides another

phonological exception to the obligatoriness of fronting wh-phrases. (52) is an example of MWF in

Romanian.

(52) Cine unde ce a adus?

who where what has brought

‘Who brought what where?’

Like SC, Bulgarian, and Russian, Romanian obligatorily fronts all wh-phrases, including wh-phrases

in echo questions. Thus, according to Comorovski (1996), (53) is bad even as an echo question.30

(53) *Ion a adus ce?

Ion has brought what

Interestingly, Comorovski notes that exceptionally, echo wh-phrases have to stay in situ in questions

that require a question as an answer. ((54b) would be unacceptable as a true, non-echo question.)

(54) a. Q: Cine a uitat s+ deschid+ para, uta?

who has forgotten to open the-parachute

b. Echo Q: Cine a uitat s- deschid- ce (anume)?

who has forgotten to open what exactly

Comorovski shows that we are dealing with a PF effect. She shows that it is impossible to assign a

proper melodic contour to (54b) if the echo wh-phrase is fronted. True questions in Romanian have

21

a melodic peak on the wh-phrase, which is immediately followed by a falli ng contour. The

intonation cannot start falli ng immediately after the true question wh-phrase if it were immediately

followed by an echo wh-phrase, echo wh-phrases being pronounced with a sharp rise in pitch.

Comorovski (p. 63) shows that a proper melodic contour can be assigned if the echo wh-phrase is

pronounced in situ.

How can we instantiate this formally?(53) shows that, as in SC, in Romanian echo wh-

phrases must front in the syntax. The same then holds for the echo wh-phrase in (54). Ignoring the

copy left by fronting the first wh-phrase, (54b) has the SS in (55a). As discussed above, if the head

of the chain created by the fronting of the echo wh-phrase is pronounced the construction cannot be

assigned a proper melodic contour, resulting in a PF violation. However, the violation can be

avoided if the tail of the chain is pronounced (55b). The construction can then be assigned a proper

intonation pattern.

(55) a. true-wh echo-whi .......verb echo-whi

b. true-wh echo-whi .......verb echo-whi

We also explain why the second wh-phrase has to be fronted on the non-echo reading. Since on this

reading the second wh-phrase is not pronounced with a sharply raised pitch the PF problem that

arises on the echo question reading of the second wh-phrase does not arise on the non-echo reading.

PF then does not license lower pronunciation of the second wh-phrase on the non-echo reading, as

it does on the echo reading. Lower pronunciation is then disallowed.

Chomsky’s (1995) Move F Hypothesis provides an alternative analysis. SC (37) and

Romanian (54b) can be analyzed as involving overt feature movement of the second wh-phrase (it

would take place in the same cycle as the movement of the first wh-phrase), leaving phonological

features of the second wh-phrase behind. The second wh-phrase then has to be pronounced in the

tail of the chain.31

(56) [FP . ta FF(/ ta)i [uslovljava / tai]]

what what conditions what

The analysis is inconsistent with Chomsky’s (1995) system, where separating FF from phonological

22

features is assumed to lead to a PF crash. Pesetsky (2000), however, argues against this position.

According to Pesetsky, there is nothing inherent to PF that would prevent feature movement prior

to spell -out.32 To make the Move F analysis work in the constructions in question we have to assume

that full phrasal movement is preferable to feature movement, at least prior to spell -out. This can be

assumed to hold generally or only in the constructions in question. Taking the latter tack would

essentially mean assuming that each movement is arbitrarily specified as either affecting or not

affecting phonological features, as in Bobalji k (1995), Groat and O’Neil (1995), and Pesetsky (1997,

1998). We further need to assume that this holds only up to convergence. The specification can be

overridden if necessary for PF convergence, as in the cases under consideration. Alternatively, we

can assume that phrasal movement is generally preferred to feature movement at least in overt

syntax. We would then be following Chomsky’s (1995). However, we cannot use Chomsky’s exact

reasoning since it does not allow for the up-to-the-PF-convergence exception to the obligatoriness

of full phrasal movement: it always forces full phrasal movement prior to spell-out. We need a

system in which phrasal movement is only a preference. A proposal by Norvin Richards, discussed

in Pesetsky (2000), achieves this. Following Richards, Pesetsky observes that taking the idea of

Attract Closest seriously would make phrasal movement more economical than feature movement

because the phrase is always the closest element with the relevant feature to the target (see also

Fukui 1997). Suppose we are allowed to look inside the closest candidate for attraction if necessary

for PF convergence. (I am departing here from Chomsky’s 1995 view of Attract Closest.) This is

exactly what would happen in the cases under consideration, where full phrasal focus movement of

the second wh-phrase results in a PF crash. Feature movement then takes place instead of full phrasal

movement.

Consider (40) and (46) under this analysis. Recall that the order of wh-phrases reflects the

order of movement to SpecCP. The highest wh-phrase koj moves first, the order of movement of

kogo and na kogo being free. In all constructions koj moves first via phrasal movement. Either na

kogo or kogo moves second. In (57a) na kogo moves second and in (57b) kogo does. Since at this

point nothing goes wrong as a result of these movements, the movements can be, hence must be,

phrasal. The first two wh-phrases are then pronounced in the raised positions.33(I use traces here for

ease of exposition)

23

(57) a.Koji na kogoj e ti pokazal tj kogo?

who to whom is shown whom

b. Koji kogoj e ti pokazal na kogo tj?

Finally, the third wh-phrase moves. In (57b) nothing goes wrong if it undergoes phrasal movement,

which is then the only option. Since the movement carries phonological features, this wh-phrase is

also pronounced in the raised position, giving us (40b). However, if the third wh-phrase undergoes

phrasal movement in (57a) the constraint against consecutive homophonous wh-phrases is violated.

To avoid this, the third wh-phrase undergoes feature movement. This wh-phrase is then pronounced

in its base-generated position, giving us (46a). Neither derivation can yield (46b), a desirable result.

Let us see if we can tease apart the Move F and the pronounce a copy analysis (PCA). Note first

that the PCA may be conceptually more appealing. The Move F analysis involves some globality

(we sometimes do Move F instead of phrasal movement in the syntax for PF reasons), which is not

the case with the PCA.34 Let us, however, see if the analyses can be teased apart empirically.

Under the most natural interpretation of the PCA we would expect successive cyclic movement

to have a PF reflex in the constructions under consideration. Unless we specifically stipulate that

only the head or the very tail of a chain can be pronounced (see Franks 1998 for a different proposal)

it seems that the second 0 ta in the SC what what construction and the echo wh-phrase in the

Romanian construction would not have to be pronounced in their base positions. This is not the case

necessarily under the Move F analysis. In fact, unless additional assumptions are adopted (for

relevant discussion see Cheng 2000, who proposes that Move F can be launched in the middle of

successive cyclic phrasal movement), under this analysis we would expect the relevant wh-phrases

to occur in the position they occupy prior to wh-movement. The test in question cannot be run for

the SC what what construction due to an interfering factor. As Bo1 kovi 2 (1997a) shows, SC has

more than one position for focus licensing of wh-phrases, as a result of which it is diff icult to

determine in more complicated constructions whether we are dealing with pronunciation of a copy

of the second what, or the head of the focus movement chain of the second what. The same problem

arises with Romanian echo wh-constructions since Romanian seems to have more than one position

where moving echo wh-phrases can be licensed. The problem, however, does not arise in Bulgarian

24

and Romanian what what constructions since, as Rudin (1988) shows (see also (29)-(30)), in these

languages only interrogative C can license non-wh-fronting of non-echo wh-phrases.35

Unfortunately, the data are not clear. (Only

one copy of the second what is pronounced. Note that (58a) and (59a) differ from (58b) and (59b),

where the indicated pronunciation is the only possibility. % indicates variation in judgments.)36

(58) a. Kakvo (*kakvo) misli (*kakvo) Ivan (%kakvo) 3 e (kakvo) obuslavlja (kakvo)?

what what thinks Ivan that conditions

‘What does Ivan think conditions what?’

b. Koj kakvo misli Ivan 3 e obuslavlja?

who what thinks Ivan that conditions

‘Who does Ivan think conditions what?’

(59) a. Ce (*ce) crede (*ce) Ion (*ce) c4 (%ce) a (*ce) determinat (ce)?

what what thinks Ion that has determined

‘What does Ion think determined what?’

b. Cine ce crede Ion c4 a determinat?

who what thinks Ion that has determined

‘Who does Ivan think determined what?’

The most plausible candidate for an intermediate landing site seems to be the embedded SpecCP.

The pre-verbal copy in (58a) can be located in the Case-checking position of what, given that, as

shown in Bo5 kovi 6 (1997b), accusative wh-phrases pass through their Case-checking position on

their way to SpecCP. So, the only unambiguous intermediate copy of wh-fronting itself is the one

immediately preceding 7 e/c8 . The judgments of my informants differ concerning the possibilit y of

pronouncing the second what in this position, most of them rejecting it. However, several interfering

factors prevent us from drawing a strong conclusion from this. First, something like a doubly fill ed

Comp filter can be an interfering factor here. Notice also that at least in some cases, Bulgarian and

Romanian are not sensitive to the wh-island constraint, which can be interpreted as indicating that

Bulgarian and Romanian wh-phrases do not have to stop in SpecCP, another interfering factor.37

Furthermore, Richards (1997) claims that movement of the second wh-phrase is not sensitive to

25

Subjacency. According to him, the first wh-phrase satisfies Subjacency with respect to the matrix

C in the constructions in question. Given his Principle of Minimal Compliance, the gist of which is

that every requirement needs to be satisfied only once, the second wh-phrase does not have to satisfy

Subjacency. Its movement can then proceed in one fell swoop. In fact, if we assume that successive

cyclic movement takes place to satisfy Subjacency the second wh-phrase in the above constructions

cannot undergo successive cyclic movement; it has to move in one fell swoop.

I turn now to an argument for the current analysis that can also help us tease apart the PCA

and the Move F analysis. Under the current analysis, the wh-phrase in situ in the constructions in

question undergoes movement in overt syntax, either full phrasal movement, as in the PCA, or

feature movement, as in the Move F analysis.As a result, we would expect the wh-phrase to be able

to license other elements from the putative raised position. One relevant phenomenon is parasitic gap

(PG) licensing. Since Bulgarian and SC do not allow PGs I focus on Romanian.38Consider (60)-

(61).39

(60) a. Cine a citit CE f9 r 9 s9 claseze?

who has read what without subj.part. files

‘Who read what without filing?

b. Ce precede ce f9 r 9 s9 influen: eze?

what precedes what without subj.part. influences

‘What precedes what without influencing?’

(61) cf. *Cine a citit cartea f; r ; s; claseze?

who has read the-book without subj.part. files

‘Who read the book without filing?’

The fact that a wh-in-situ can license a PG provides strong evidence for the current approach, on

which the wh-in-situ undergoes movement in overt syntax in spite of being pronounced in situ.

Notice that (60a-b) contrast in the relevant respect with their English counterparts.

(62) a. *Who read WHAT without filing?

b. *What precedes what without influencing?

26

This is not surprising under the current analysis, since (60) and (62) receive different analyses in

spite of the superficial similarity. This is particularly clear under the PCA, where the wh-in-situ in

(60) undergoes phrasal movement in overt syntax that does not differ syntactically in any relevant

respect from wh-movement of what in (63). It is then not surprising that (60) patterns with (63) and

not (62).

(63) What did John file without reading?

Under the Move F analysis of (60), we have to assume that formal features suff ice for PG licensing.

(Only the formal features of the wh-in-situ move, semantic and phonological features stay behind.)

Furthermore, to account for the contrast between (60) and (62) we need to assume that the in situ wh-

phrase in English (62) does not move in LF.40 If it were to move in LF, in Chomsky’s (1995) system

it would undergo feature movement (see, however, Pesetsky 2000), li ke the in situ wh-phrase in (60)

under the Move F analysis of these constructions. True, feature movements in (60) and (62) could

be taking place in different components, overt syntax and LF. (This would not be the case in systems

which dispense with LF.) However, this should be irrelevant as long as we do not assume that PG

licensing is an SS phenomenon, which would be inconsistent with the Minimalist Program. If either

of the two assumptions necessary to make the Move F analysis of (60) work cannot be maintained

we have here an argument for the superiority of the PCA over the Move F analysis.

The PG data show that we are dealing here with a new type of in situ wh-phrases not attested

in English-type, French-type, and "true" wh-in-situ languages. We have already seen that, in contrast

to in situ wh-phrases in MWF languages, in situ wh-phrases in English multiple questions cannot

license PGs. The same holds for in situ wh-phrases in French and Malay wh-in-situ questions.41

(64) a. *Il a lu quoi sans classer? (French)

he has read what without to-file

‘What has he read without filing?’

b. cf. Qu’a-t-il lu sans classer?

c. *Kamu aturkan buku yang mana tanpa baca? (Malay)

you filed book that which without reading

27

‘Which book did you file without reading?’

d. cf. Buku yang mana kamu aturkan tanpa baca?

book that which you filed without reading

This means that whatever analysis of (60) is adopted it should not be applied to (64). So, if the in

situ wh-phrases in (60) undergo feature movement, the in situ wh-phrases in (64) cannot be

undergoing feature movement.42 And if the in situ wh-phrases in (60) undergo phrasal movement

with pronunciation of a lower copy, the same should not hold for the in situ wh-phrases in (64).43

Either way, we are dealing here with a distinct type of in situ wh-phrases, different from in situ wh-

phrases in non-MWF languages like English, Malay, and French. As a result, determining the most

adequate analysis of (60) has important ramifications for analyzing in situ wh-phrases in non-MWF

languages.

2.3 A syntactic exception to the obligatoriness of wh-fronting in MWF languages

Comorovski (1996)notes that echo wh-phrases in Romanian can exceptionally stay in situ within

non-Relativized Minimality islands (i.e. non-wh-islands). (65) contrasts with (53) on the echo

question reading. Notice that overt wh-movement out of the island is not allowed regardless of the

reading.44

(65) Ion a auzit zvonul c< Petru a cump< rat CE?

Ion has heard the rumor that Peter has bought what

(66) *Ce a auzit Ion zvonul c< Petru a cump< rat?

Assuming that islandhood is syntactic in nature we are dealing here with a syntactic exception to the

obligatoriness of MWF in Romanian. The exception is readily accounted for under the Move F

analysis, on which Romanian wh-phrases undergo Move F if phrasal movement is not possible. Ochi

(1998) and Agbayani (1998) argue that phrasal movement but not feature movement is subject to

non-Relativized Minimality-type islands. According to them, feature movement is subject only to

Relativized Minimality islands through Attract Closest. If they are right, full phrasal movement of

the echo wh-phrase in (65) is not possible. The wh-phrase can then undergo feature movement.45 The

Move F analysis enables us to account for (65) while still having the wh-phrase undergo movement

28

to C, which is desirable given the ungrammaticality of (53).Given that (65) is unacceptable on the

true question reading a question arises why the Move F derivation is unavailable on this reading. I

speculate that Cheng’s (1997) clausal typing is the culprit. A whole wh-phrase, including its

phonological features, must be present in SpecCP in overt syntax in Romanian to type a clause as

a question. This rules out the possibility of wh-in-situ on the true question reading in Romanian.

The V-2 effect raises a potential problem for the Move F analysis. Normally, in both subject

and non-subject questions verbal elements occur in the second position on both the echo and the non-

echo reading of the fronted wh-phrase. As a result, they precede the subject in non-subject questions.

(67) Ce a spus M= d= lina?

what has said Madalina

‘What did Madalina say?’

Under the Move F analysis, (65) is a non-subject question with the verbal elements following the

subject. This is not a problem if the V-2 effect is phonological in nature, as suggested in Bo> kovi ?(2001), Chomsky (1995), and Rice and Svenonius (1998), among others, for various languages.

Alternatively, we can assume that the subject in (65) is located in a pre-SpecCP topic position.

Turning to the PCA, under this analysis (65) involves phrasal movement of the echo wh-

phrase in overt syntax. The head of the chain created by the movement is deleted and the copy is

pronounced. (Deletion of the head has to be sanctioned by PF reasons, which are discussed below.)

(68) Ce ... [NP ...ce ]Under this analysis (66) and (65) have the same syntactic derivation. As a result, accounting for the

contrast between them becomes diff icult. The only way to preserve the PCA seems to be to assume

islandhood is at least to some extent a PF property. Some older approaches to islandhood do assume

this, e.g., Perlmutter (1972), revived in a slightly different form in Pesetsky (1997,1998). (For PF

approaches to islandhood, see also Lasnik 2000 and Merchant 1999.) For Perlmutter, syntactic

movement is not constrained by islands. What is constrained by islands is the obligatory deletion of

the copy left by movement. The deletion fails when an island intervenes between the head of a chain

and its copy. Interpreting this as a PF violation leads us to pronounce a copy instead of the head of

29

the chain under Franks’ approach to pronunciation of chains. The PCA seems to be based on a rather

unorthodox view of locality restrictions. This is actually not true. The analysis is based on the more

or less standard view of the saving effect of resumptives with respect to locality of movement (but

see Boeckx 2001), which implies that at least to some extent, locality is a PF phenomenon. It is well -

known (see Shlonsky 1992 and Pesetsky 1997,1998 among others)that in a number of languages,

e.g. Hebrew, Arabic, Irish, and English, a locality violation can be saved by phonologically realizing

a copy within the island as a resumptive pronoun. Using a resumptive in these languages is a last

resort strategy employed only when movement would otherwise violate locality restrictions on

movement.

(69) a. *There is one worker who the company fired the employee that had treated badly.

b. There is one worker who the company fired the employee that treated him badly.

c. *This is the guy who I like him. (Pesetsky 1998:364)

Apparently, phonologically realizing a copy within an island can rescue a construction from a

locality violation.46 This, I propose, is what happens in (65). Movement out of the island takes place.

The construction is saved from a locality violation by phonologically realizing a copy within the

island. (Note that the typing requirement is irrelevant since we are not dealing with a true question.)

The only difference is that in (69b) the copy is realized as a resumptive pronoun and in (65) the full

copy is pronounced. Pesetsky (1997,1998) proposes that in (69b) the tail of the chain is pronounced

as a pronoun due to a constraint that requires copies that are not heads of chains to be as close to

unpronounced as possible. Pronunciation of @ -features, i.e. pronominal pronunciation, is the

minimal pronunciation. The resumptive pronoun strategy cannot be employed in (65) because quite

generally, echo wh-phrases cannot be associated with resumptive pronouns.47 Since a resumptive

pronoun is not an option, a full copy is pronounced. Why can’ t we phonologically realize both the

head and the tail of the wh-chain in (65)? Nunes (1999) argues that pronouncing both the head and

the tail of a chain would violate Kayne’s (1994) Linear Correspondence Axiom (LCA) so that the

chain could not be linearized. Nunes considers the head and the tail of a chain to be non-distinct for

the purposes of the LCA. As a result, realizing both phonologically would result in a conflicting

ordering. Suppose we decide to delete neither the head nor the tail of the chain created by movement

30

of ce in (65). Given the LCA, the wh-phrase will have to both precede (because of ce in SpecCP) and

follow (because of ce in the base-generated position) other words in the sentence. Linearization

therefore fails. What about the resumptive pronoun examples? Why are they not violating the LCA?

It seems plausible that the wh-phrase and the resumptive pronoun are not non-distinct for the

purposes of the LCA since they do not receive the same phonological realization. The LCA is then

not violated in (69b).

Let me finally point out that Franks’ and Pesetsky’s approach to PF realization of chains are

very similar. Forcing a copy that is not the head of a chain to be as close to unpronounced as possible

entails that if there is no reason to pronounce it, it will not be pronounced. For Franks, and the same

seems to hold for Pesetsky, the relevant reasons are phonological, which makes sense given that

copy pronunciation is a PF phenomenon. In principle, the head of a chain can always be pronounced.

Whenever copies (by copies I mean everything but the head of a chain) are all deleted the head of

the chain must be pronounced to avoid violating Recoverabilit y of Deletion. When a copy must be

fully realized phonologically for independent reasons, the head must be deleted to avoid violating

the LCA. With partial phonological realization of a copy, as with resumptive pronouns, the head of

the chain cannot be deleted. Its deletion would violate Recoverabilit y of Deletion - a wh-phrase and

a pronoun obviously cannot be considered non-distinct for the purposes of Recoverabilit y of

Deletion.

To sum up,phonological and syntactic exceptions to the obligatoriness of wh-fronting in

MWF languages can be accounted for by adopting either the PCA or the Move F analysis. (Under

the PCA there are no purely syntactic exceptions.) At this point, we cannot completely conclusively

choose one of the two analyses. The PG data, however, do favor the PCA.48 Determining which

analyses is more adequate has important consequences for analyzing in situ wh-phrases in non-MWF

languages.

3 Conclusion

I showed that MWF languages do not display uniform behavior with respect to wh-movement, thus

eliminating this type of language from the crosslinguistic typology concerning the behavior of wh-

phrases with respect to wh-movement in multiple questions. This leaves us with 3 types, represented

by English, French, and Chinese. MWF languages are scattered across these 3 types: Bulgarian is

31

a MWF counterpart of English, SC of French, and Russian of Chinese. The behavior of MWF

languages with respect to wh-movement is camouflaged by the focus requirement, which forces all

wh-phrases to move overtly independently of wh-movement. We have seen that there are 3 classes

of exceptions to the obligatoriness of wh-fronting in MWF languages: semantic, phonological, and

syntactic. The semantic exceptions are explained away by the focus nature of the additional

movement of wh-phrases in MWF languages. I have considered two explanations for phonological

and syntactic exceptions: one based on the Move F Hypothesis and one based on the possibilit y of

pronunciation of lower copies of non-trivial chains sanctioned by PF considerations. The latter

provides evidence for the copy theory of movement. The exceptions to the obligatoriness of wh-

fronting have led me to posit a new type of in situ wh-phrases, distinct from in situ wh-phrases in

non-MWF languages.

Notes*Parts of this paper were presented at Formal Approaches to Slavic Linguistics 5 (Wabash College),

Nanzan University Syntax Workshop, LSRL 27 (University of Cali fornia at Irvine), WCCFL 16

(University of Washington), Comparative Slavic Morphosyntax Workshop (Indiana University),

Generative Grammar Workshop of the Basque Center for Language Research, Formal Approaches

to South Slavic and Balkan Languages 3 (University of Plovdiv), NELS 30 (Rutgers), and colloquia

at Harvard, Kanda University of International Studies, New York University, Princeton, University

of Toronto, Cornell , Université de Paris 8, University of Maryland, a course at the 5th European

Summer School in Generative Grammar, and seminars at the University of Connecticut. I thank all

these audiences for thought provoking questions. For valuable comments, special thanks are due to

Wayles Browne, Noam Chomsky, Steven Franks, Norbert Hornstein, Richard Kayne, Masao Ochi,

David Pesetsky, Charles Reiss, LI reviewers, and especially Cedric Boeckx, Howard Lasnik, and

Sandra Stjepanovi A . For help with judgments, I am especially grateful to Larisa Avram, Michèle

Bacholle, Sigrid Beck, Cedric Boeckx, Gabriela Bulancea, Ileana Comorovski, Alexandra

Cornilescu, Viviane Déprez, Alexander Grosu, C.-T. James Huang, Dana Isac, Roumyana Izvorski,

Mariana Lambova, Géraldine Legendre, Kalaivahni Muthiah, Masao Ochi, Suba Rangaswami,

Mamoru Saito, Penka Stateva, Arthur Stepanov, Sandra Stjepanovi A , and SaB a Vuki A . This research

was partially supported by a grant from the International Research and Exchanges Board, with funds

32

1.Malay may be a better example of a wh-in-situ language since like Japanese (see Watanabe 1992),

Chinese has been argued to involve overt null operator movement to SpecCP in questions (see Cole

and Hermon 1995). Cole and Hermon show that the null operator analysis does not work for Malay.

2.See Rudin (1988), BoC kovi D (1997b, 1998b, 1999), Richards (1997, 1998), and Pesetsky (2000).

One argument that the fixed order of wh-phrases in Bulgarian (ia-b) is a result of Superiority

concerns the fact that (ib) improves with D-linked and echo wh-phrases. (Koj in (id) is an echo wh-

phrase.) The same happens with Superiority violations in English (ii ). All the above authors argue

that the wh-phrase that is first in the linear order in Bulgarian is the one that moves first. The second

wh-phrase either right-adjoins to the first wh-phrase, located in SpecCP, as in Rudin (1988), or

moves to a lower SpecCP (the first wh-phrase being located in the higher SpecCP), as in Richards

(1997, 1998) and Pesetsky (2000). For another approach to Bulgarian MWF, see Grewendorf (2001)

and Kim (1997).

(i) a. Koj kakvo e kupil?

who what is bought

‘Who bought what?’

b. *Kakvo koj e kupil?

c. ?Koja kniga koj E ovek e kupil?

which book which man is bought

‘Which man bought which book?’

d. ?Kakvo KOJ e kupil?

(ii) a. *What did who buy?

b. Which book did which man buy?

c. What did WHO buy?

3.I ignore the echo reading. I do not give indirect questions due to an interfering factor. Since they

formally do not differ from root questions they can be easily analyzed as root questions with the root

clause treated as an adsentential. The problem does not arise with correlatives (7) and existentials

(8), which also contain embedded questions (see Izvorski 1996, 1998). In BoC kovi D (1997c) I show

provided by the US Department of State (Title VIII program) and the National Endowment for

Humanities.

33

that when the interfering factor is controlled for, indirect questions also show Superiority effects.

4.There is some variation with respect to the relevant data. Ljilj ana Progovac (p.c.) informs me that

in her judgment, SC patterns with Bulgarian. Léa Nash (p.c.) informs me that for her, Russian

patterns with SC. (This holds for both Superiority and the data concerning the interpretation of

multiple questions discussed below, which provides strong evidence for the current analysis.)

5.I am considering only the reading on which the echo question asks for repetition of what the

questioner has not heard (see section 2.1 for another echo question reading). SC F ta Ivan kupuje is

thus ambiguous between the echo and the non-echo reading. Notice that one of my Bulgarian

informants accepts (14a) and (16b). However, even for this speaker, echo wh-phrases in situ are

better on the surprise echo question reading than on the request for repetition reading, in line with

the discussion in section 2.1 below (see this section for discussion of different echo readings).

6.Contrastive focus, also referred to as identificational or narrow focus, expresses exhaustive

identification and is accompanied by emphatic stress. It is important to distinguish it from simple

new information focus, also referred to as wide or presentational focus.

7.Not all Slavic speakers obligatorily front focused non-wh-phrases. All my informants have this

option. However, a few of them can also leave focused non-wh-phrases in situ. Most Slavic speakers

have at least a strong preference for fronting them, see, e.g., King 1993:105, who claims that this is

a strong tendency in Russian. (Stepanov 1998:461 claims such phrases must move in Russian.) We

can decide to ignore the optional fronting pattern, since it is clearly dispreferred, perhaps for all

speakers. If we don’t, there are two ways to account for it. (I refer to it as Variety I.) We can posit

a minor difference in the lexical specification of wh-phrases and focused non-wh-phrases by

assuming that wh-phrases have a strong focus feature while focused non-wh-phrases can have either

a strong or a weak focus feature in Variety I (see Kidwai 1999 for a similar proposal for Hindi/Urdu

and Malayalam). There is a more principled alternative. There is an interesting similarity in the

behavior of different types of wh-phrases in Malay and different types of focalized elements in

Variety I. Malay argumental wh-phrases can either move or stay in situ. Cole and Hermon (1995)

show that Malay argumental wh-in-situ does not involve null operator movement, as argued for

Japanese by Watanabe (1992). They show that no wh-movement of any kind takes place in Malay

argumental wh-in-situ. On the other hand, adjunct wh-phrases must move overtly. The gist of their

34

analysis is that although in principle wh-movement is optional in Malay, adjuncts must move

because they are uninterpretable in situ. The analysis can be applied to focus in Variety I. Suppose

focus movement is in principle optional in Variety I. (The optionality can be a result of different

lexical choices, as in Cole and Hermon’s analysis.) Suppose furthermore that in MWF languages wh-

phrases are interpretable only in the focus position. As a result, wh-phrases must undergo focus

movement even in Variety I, where focus movement is in principle optional, since they would

otherwise be uninterpretable.

8.Focus movement is insensitive to Superiority. BoG kovi H (1998b,1999), who provides an

explanation of this fact, shows that this holds for Bulgarian as well as SC and Russian (see section

2.2).Slavic languages seem to differ arbitrarily regarding which elements license focus. This is not

surprising since it is well -known that there is considerable crosslinguistic variation regarding where

focus is li censed under movement (see, e.g., Kiss 1995, especially p. 23). Focusing on Slavic,

BoG kovi H (1997a) argues that in Bulgarian, the focus licensor is interrogative C, in SC Agr and in

some cases interrogative C. Izvorski (1993) and Lambova (in press) argue for a separate focus-

licensing head below C for Bulgarian. Stjepanovi H (1999b) gives an analysis of SC with Agrs and

Pred as focus licensors. Stepanov (1998) argues that the focus licensor in Russian is Agrs and King

(1993) I .

9.For explanation of this, see BoJ kovi K (in press). Under the analysis given in this work, which is

based on Hagstrom’s (1998) semantics of questions, languages that have obligatory overt movement

of a wh-phrase to SpecCP cannot license single-pair answers, while languages that do not have it

may, but do not have to, allow such answers. In other words, not filli ng SpecCP by a wh-phrase

overtly is necessary but not sufficient for licensing single-pair answers.

10.Unfortunately, no definite conclusion can be drawn from examination of contexts where SC must

have overt wh-movement. The relevant test either cannot be run due to interfering factors concerning

the interpretation of relevant constructions (this holds for correlative, existential, and li-

constructions) or it fails to give a clear result due to the murkiness of judgments (long-distance

multiple questions, which are not very productive to start with). (The li counterpart of (18), Ko li j eLta kupio, cannot be used in the situation depicted with respect to (18). However, I hesitate to draw

a strong conclusion from this since the li -construction is not a "neutral" question semantically.) See,

35

however, the discussion of TC in section 2.1, which provides additional evidence for the claim made

in this section.

11.This was noted in Wachowicz (1974) and Pesetsky (1987, 1989), who give Polish and Russian

examples. (Pesetsky also mentions Czech and Romanian.) In (13)-(15) I have used wh-phrases that

are more diff icult to D-link. Pesetsky and Wachowicz observe that non-inherently D-linked wh-

phrases can stay in situ when used in a context forcing a D-linked interpretation(those that in

principle can be D-linked,see Pesetsky 1987:127for an exception),which holds for all the languages

considered. Throughout the paper I assume non-D-linked contexts for non-inherently D-linked wh-

phrases.

12.See also Reinhart (1997:158), who says that "D-linked constituents are not particularly good

foci." Pollock et al (1998) give data from Bellunese which seem to indicate that in this language, D-

linked and non-D-linked wh-phrases appear in different positions. This can be accounted for if non-

D-linked wh-phrases are focalized in this language and if D-linked wh-phrases cannot occur in a

focus position.

13.Je is a second position cliti c. SC second position cliti cization is a murky phenomenon that may

involve PF word reordering (see BoM kovi N 2001 and Franks and King 2000), hence I ignore it here.

14.What is important for us is that the movement in question is not focus movement, which is clear

given the contrast in (30)-(31). I use the word scrambling merely to distinguish the movement in

question from focus and wh-movement and indicate the appearance of optionality. Whether we are

dealing with true optionality (i.e. whether there is truly optional scrambling) remains to be seen.

15.For Rudin, this as an argument that the wh-phrases are a constituent. This is true under the

adjunction to SpecCP analysis, but not under the multiple specifiers analysis. Under this analysis,

(30) can be ruled out due to a feature clash:a [-wh,-focus] element is located in a [+wh,+focus] CP.

Note that (30) improves markedly if the parenthetical is contrastively focused (see BoM kovi N 1998c),

which is not surprising under the focus movement analysis. In fact, this is an argument for the

analysis.

16.Cheng leaves open how the typing is carried out with French wh-in-situ. I have nothing new to

add concerning French. For relevant discussion, see Boeckx (1999) and Cheng and Rooryck (2000).

17.I am departing from Cheng in the technical aspect of the typing analysis, maintaining its spirit.

36

18.In BoO kovi P (1998a, 2000) I apply the LF C-insertion analysis to French. LF C-insertion results

in wh-in-situ, i.e. lack of overt wh-movement, in French. As in SC, overt C-insertion triggers overt

wh-movement. (This is where French differs from Chinese. In Chinese no wh-movement needs to

take place overtly even when C is inserted overtly. More formally, the +wh-feature of C is weak in

Chinese and strong in French. French allows wh-in-situ because C can be in certain contexts inserted

in LF, which is a possibility under Chomsky’s 1995, chapter 4 definition of strength.)

19.They precede wh-phrases in SpecCP. Note that (33) can contain a wh-phrase in the highest phrase

so that clausal typing is not a problem. TCs are, however, often treated as extrasentential and ignored

for clause-internal requirements. Thus, Q avar and Wilder (1999) and Schütze (1994), who adjoin

TCs to CP, treat TCs as extrasentential (i.e. as not belonging to the same clause as elements

dominated by CP) for the purpose of cliti c placement. It is then possible that the presence of a TC,

which is not dominated by CP hence can be said not to make the CP phonologically realized, does

not force clausal typing within CP in (33). Wh-movement still must take place in (33) for reasons

noted above. Note that even if we assume that TCs are located in the Spec of a head taking CP as

complement, as in Tomi R (1996), wh-movement wil l still be forced in (33). Under this analysis it

is also natural to ignore the projection hosting TCs for clausal typing purposes.Since according to

TomiR the projection hosts only elements denoting old information, wh-phrases can never move to

it, i.e. they always remain below it. Under this analysis, we can assume that clausal typing takes

place up to CP, i.e. that CP closes its domain. I assume that for one of the above reasons, TCs do not

affect clausal typing.

20.The focus requirement can in fact be considered the defining characteristic of MWF languages.

21.The exception was pointed out to me by Wayles Browne (p.c.).

22.It can be marginally fronted if very heavily stressed. With neutral stress, * S ta S ta uslovljava is

bad.

23.One of my Bulgarian informants does not have the constraint. All others in all four languages do.

Similar constraints are found in other languages, e.g. Italian (Napoli ’s 1976 constraint on cliti cs),

Turkish (Kornfilt ’s 1986 constraint on consecutive sequences of compound and possessive markers),

Japanese (some instances of the Case-marker drop), and Ancient Greek (the ban on homophonous

sequences of articles, see Golston 1995). SC has this kind of effect in other constructions as well .

37

As shown in (i), the accusative feminine cliti c je is replaced by ju when adjacent to the auxili ary je

‘is’.

(i) a. Oni su je/*ju zaboravili.

they are her forgotten

‘They forgot her.’

b. On ju/*je je zaboravio.

he her is forgotten

Howard Lasnik (p.c.) notes that the possessive of boys must be boys’ [boyz] and not boys’s [boyziz]

though the relevant phonetic sequence is possible, as in the family name the Boys’s. This shows that

we are dealing with a morphological rather than a phonetic effect. (Note that the antihomophony

effect is not universal. In fact, there are exceptions to it even in the languages cited above.)

24.As argued in BoT kovi U (1997b), the object wh-phrase checks the strong +wh-feature of C in (41)

rather than the adjunct because it moves to its Case-checking position prior to wh-movement, thus

ending up higher than the adjunct prior to wh-movement. Notice also that (i) shows that (41)-(44)

do not exhibit the same phenomenon as English (ii ), where addition of a lower wh-phrase rescues

a Superiority violation (see Pesetsky 2000 for a recent discussion of this effect.)

(i) a. *Kogo koj kak e tselunal?

b. *Kogo koj kakvo e pital?

(ii) a. *What did who buy?

b. (?)What did who buy where?

25.The explanation is too involved to repeat here. Richards (1997) gives a non-unified analysis of

Superiority in Bulgarian and SC which accounts for the Bulgarian paradigm. However, it is shown

in BoT kovi U (1998c) that the analysis of SC does not work. (It does not account for the full paradigm

and is based on incorrect assumptions about the syntax of SC.)Also, it cannot be extended to

Russian, which has weak crossover effects with clause internal wh-fronting, a problem for Richards’

analysis.

26.Whether SC has the same structure as Bulgarian when it must have wh-movement is unclear.

BoT kovi U (1997a) analyzes SC constructions of this type differently from Bulgarian. However, this

may be wrong. The main argument against treating any SC question on a par with Bulgarian

38

concerns wh-islands. However, it is shown in BoV kovi W (1998c, in preparation) that the wh-island

test is irrelevant. It is worth noting here that wh-phrases are more diff icult to split by a parenthetical

in SC constructions that must involve overt wh-movement, which is a characteristic of Bulgarian

questions (see (30)). (I ignore the li and the correlative construction, since the relevant parenthetical

placement is ruled out in these constructions for independent reasons. Notice also that, li ke Bulgarian

MWF constructions, SC (ic,e,g) improve if the parenthetical is contrastively focused.)

(i) a. Ko, po tebi, koga tuX e?

who according you whom beats

‘Who, according to you, is beating whom?’

b. ?Ko koga, po tebi, tuX e?

c. *Ko, po tebi, koga vjeruju da tuX e?

who according you whom believe that beat

‘Who, according to you, they believe is beating whom?

d. ?Ko koga, po tebi, vjeruju da tuX e?

e. ??Tom X ovjeku, ko, po tebi, V ta poklanja?

that man.dat who according you what bestows

‘To that man, who, according to you, bestows what?’

f. Tom X ovjeku, ko V ta, po tebi, poklanja?

g. *Ima ko, po tebi, V ta da mu proda.

has who according you what part. him sells

‘There is someone who, according to you, can sell him something.’

h. ?Ima ko V ta, po tebi, da mu proda.

(ic-h) indicate that the SC constructions in question should be treated on a par with Bulgarian MWF.

27.Chomsky’s analysis is slightly more complicated. He also argues there is a preference for deletion

in the head of operator-variable chains, motivated by *John wondered which picture of Tomi hei

liked.

28.Pesetsky’s (1997,1998) system, which antecedes Franks’ , is very similar to it (see also Bobalji k

1995, Hiramatsu 1997, and BoV kovi W 2001). However, Pesetsky does not explicitl y make the claim

that only PF considerations can license lower pronunciation, a position I wish to maintain.

39

29.Although SC is a free word order language, when the subject and the object cannot be

disambiguated through case inflection there is a strong tendency to interpret the first NP as the

subject and the second NP as the object. The same tendency exists in (49).

30.Some of my informants do not share Comorovski’s judgment. I am focusing here on the dialect

in which (53) is unacceptable as an echo question.

31.The Move F analysis seems restateable in terms of Chomsky’s (1999) Agree.

32.Note that phonological features remain together after the movement. And if PF needs formal

features, their copy is present in the same position with phonological features even after Move F.

33.I assume later movements cannot affect the locally made decision to do phrasal movement here.

34.Similar globality is quite generally present in Chomsky’s (1995) view of Move F, where we

always do phrasal movement in overt syntax for PF reasons, as well as Chomsky’s (1995, chapter

3) view of strength as an illegitimate PF object.

35.This does not hold for echo wh-phrases. This does not provide evidence that non-wh-fronting of

non-echo and echo wh-phrases in these languages are different phenomena. It is possible that though

there is more than one potential li censor for non-wh-fronting in these languages, interrogative C

must be the licensor whenever present. In BoY kovi Z (2001) I suggest that, in contrast to SC, in

Bulgarian interrogative C is always inserted overtly. The reason for this is that, in contrast to SC,

interrogative C in Bulgarian is lexically specified as a PF verbal aff ix. The requirement cannot be

satisfied if the C is not inserted overtly. Evidence for the difference between Bulgarian and SC is

provided by the fact that in Bulgarian but not in SC, the C must be V-adjacent. (Romanian patterns

with Bulgarian.)

(i) a.*Kakvo toj dade na Petko? (Bulgarian)

what he gave to Petko

‘What did he give to Petko?’

b. Kakvo dade toj na Petko?

c. [ ta on dade Ivanu? (SC)

what he gave Ivan.dat

Since in true questions interrogative C must be inserted overtly, non-echo wh-phrases must move

to interrogative CP in Romanian and Bulgarian. In pure echo questions it appears that interrogative

40

C does not have to be inserted at all. Hence, echo wh-phrases can be licensed in other positions.

36.Among my informants who have the what what constraint, both Bulgarian informants can realize

the second what before the verb and one can realize it before \ e.Only one of my Romanian

informants allows the indicated intermediate pronunciation. However, there is an interfering factor

with realizing ce before the main verb in (59a). Only certain cliti c-li ke adverbs can intervene

between the auxili ary and the participle, which suggests that the auxili ary is a verbal cliti c (see

Dobrovie-Sorin 1994:10-11).

37.For some authors (see Rudin 1988, Koizumi 1999, and Richards 1997), Bulgarian wh-phrases

move through SpecCP even in wh-island configurations, which would eliminate the interfering

factor. The analysis relates the resistance of Bulgarian to the wh-island constraint to the possibilit y

of MWF. See, however, Bo] kovi ^ (1998c, in preparation) for criti cism of this analysis. (For one

thing, I show that in all relevant respects Bulgarian behaves like Swedish, a non-MWF language.)

38.Russian allows them but interfering factors prevent us from running the PG test. (According to

Arthur Stepanov (p.c.), PGs are not natural in Russian MWF constructions.) Bulgarian and SC have

the counterparts of (60b), but I believe that in these languages such constructions should be analyzed

as involving Across-the-Board movement (ATB). (Other PG constructions from English are

unacceptable in these languages. For what it is worth, (i) gives the relevant examples from

Bulgarian.)

(i) a. ?(?)Kakvo opredelja kakvo bez da o_ akva?

what determines what without part. anticipates

‘What determines what without anticipating?’

b. *Koj opredelja kakvo bez da o_ akva?

who determines what without part. anticipates

39.(60) involve the PF exceptions to the obligatoriness of MWF. (Capitals indicate an echo wh-

phrase.) All my informants accept (60a). The judgments are divided for (60b), with the majority

accepting it. Note that there are potentially interfering factors in the PG test. PF information may

also be involved in PG licensing (see Franks 1993). This may help us account for the judgment of

speakers who do not accept (60b), given that under both the PCA and the Move F analysis the PG

licensor is not phonologically realized in the raised position. (Note that there are analyses, e.g.,

41

Nunes 2001, on which we would not necessarily expect PGs to be licensed in the constructions in

question under the current analysis. A Nunes-style analysis might be appropriate for the speakers

who do not accept (60b) as well as the ATB construction from footnote 38. (Nunes extends his

analysis of PGs to ATB.)

40.This is so even if the in situ wh-phrase is not D-linked, contra Pesetsky (1987). Note that the PCA

is consistent with Pesetsky’s claim that non-D-linked in situ wh-phrases move in LF in English.

41.The term in situ wh-phrase refers to any wh-phrase that is not pronounced in an operator position.

The terms wh-in-situ questions and wh-in-situ languages are used in the same sense. Notice that I

avoid using as ill ustration wh-in-situ languages that allow null objects, such as Chinese and

Japanese. It is difficult to tease apart PGs and null objects in these languages.

42.The Move F analysis of (60) is thus inconsistent with my (1998a, 2000) analysis of French wh-in-

situ constructions, which I argue involve feature movement based on their locality restrictions.

(Incidentally, at least some of these locality restrictions do not hold in the Romanian constructions

in question. Compare, e.g., (59a) with my observation that French disallows long-distance in-situ

questions.) Thus, to the extent that it is successful, my (1998a, 2000) analysis favors the PCA

treatment of (60) over the Move F treatment. Needless to say, if the latter turns out to be correct, it

would invalidate the Move F analysis of French wh-in-situ. Notice also that in Bo kovi a (2000) (see

also Cheng and Rooryck 2000), I provide evidence that French wh-in-situ and wh-in-situ in

traditional wh-in-situ languages like Chinese and Japanese should not be analyzed in the same way.

43.The PCA is thus inconsistent with analyzing wh-in-situ in French and Malay as involving phrasal

movement of the in situ wh-phrase that takes place prior to spell -out, with subsequent pronunciation

of a lower copy. (For analyses along these lines for wh-in-situ languages where the PG test cannot

be run (see footnote 41), see Groat and O’Neil 1996:131 and Bobalji k 1995:360. Pesetsky 2000

suggests this analysis for Chinese, and a Move F analysis for Japanese.)

44.I again focus on the dialect in which echo wh-phrases must move, where (i) contrasts with (65).

(i) *Ion crede cb Petru a cumpb rat CE?

Ion believes that Peter has bought what

Recall also that, as discussed above, there is more than one possibilit y for the landing site of echo

wh-phrases. E.g., the echo wh-phrase in (i) can either stay within the embedded clause or move to

42

the matrix clause, as illustrated in (ii). (Ion in (iia) can be a topic located outside of CP.)

(ii) a. Ion CE crede cc Petru a cumpc rat?

b. Ion crede cc CE a cumpc rat Petru?

I assume that the same options are in principle available for the echo wh-phrase in (65). As will

become clear during the discussion below, only the derivation on which the echo wh-phrase moves

syntactically into the matrix clause can yield (65), where the echo wh-phrase is pronounced in situ.

45.With wh-islands, full phrasal movement is allowed. As expected, feature movement is degraded.

(i) a. CE se întreabc Ion cine a cumpc rat?

what refl wondered Ion who has bought

‘What did Ion wonder who bought?’

b. ?(?)Ion se întreabc cine a cumpc rat CE?

46.For PCA treatments of resumptives, see Fox (1994) and Pesetsky (1997,1998). We can

implement the PF effect on locality by assuming that PF reali zation removes the star assigned to

copies due to locality violations (see Chomsky and Lasnik 1993). Note that resumptives in English

cannot occur in intermediate positions, as shown by *the worker who you recently heard the rumor

him that they had treated badly. This may be a result of more general constraints on pronoun

placement in English - resumptives can occur only in positions in which pronouns in general can

occur in the language.

47.Resumptive pronouns sound best with relative clauses. Speakers who do accept them in true

questions seem to accept them only with heavily D-linked wh-phrases (see Boeckx 2001). Thus,

English *Who did the company fire the employee that treated him badly is judged unacceptable.

48. Recall that the Move F analysis is restateable in terms of Chomsky’s (1999) Agree. The PG data

seem particularly problematic for the Agree version of this analysis.

References

Agbayani, Brian. 1998. Generalized pied-piping and island effects. In Proceedings of NELS 28, 1-

14. GLSA, University of Massachusetts, Amherst.

Billings, Loren, and Catherine Rudin. 1996. Optimality and Superiority: A new approach to overt

multiple wh-ordering. In Formal Approaches to Slavic Linguistics: The College Park Meeting,

1994, ed. Jindrich Toman, 35-60. Ann Arbor: Michigan Slavic Publications.

43

Bobaljik, Jonathan. 1995. Morphosyntax: The syntax of verbal inflection. Doctoral dissertation,

MIT, Cambridge, Mass.

Boeckx, Cedric. 1999. Properties of French interrogatives. Ms., University of Connecticut, Storrs.

Boeckx, Cedric. 2001. Mechanisms of chain formation. Doctoral dissertation, University of

Connecticut, Storrs.

Bod kovi e , f eljko. 1997a. Fronting wh-phrases in Serbo-Croatian. In Formal Approaches to Slavic

Linguistics: The Indiana Meeting, 1996, ed. Martina Lindseth and Steven Franks, 86-107. Ann

Arbor: Michigan Slavic Publications.

Bod kovi e , f eljko. 1997b. On certain violations of the Superiority Condition, AgrO, and economy

of derivation. Journal of Linguistics 33: 227-254.

Bod kovi e , f eljko. 1997c. Superiority effects with multiple wh-fronting in Serbo-Croatian. Lingua

102: 1-20.

Bod kovi e , f eljko. 1998a. LF movement and the Minimalist Program. In Proceedings of NELS 28,

43-57. GLSA, University of Massachusetts, Amherst.

Bod kovi e , f eljko. 1998b. Multiple wh-fronting and economy of derivation. In Proceedings of the

West Coast Conference on Formal Linguistics 16, 49-63. Stanford University, Stanford, Cali f.

Bod kovi e , f eljko. 1998c. Wh-movement and wh-phrases in Slavic. Position paper presented at the

Comparative Slavic Morphosyntax Workshop, Spencer, Indiana, June 1998.

Bod kovi e , f eljko. 1999. On multiple feature-checking: Multiple wh-fronting and multiple head

movement. In Working minimalism, ed. Samuel Epstein and Norbert Hornstein, 159-187.

Cambridge, Mass: MIT Press.

Bod kovi e , f eljko. 2000. Sometimes in SpecCP, sometimes in-situ. In Step by step: Essays on

minimalism in honor of Howard Lasnik, ed. Roger Martin, David Michaels, and Juan Uriagereka,

53-87. Cambridge, Mass.: MIT Press.

Bod kovi e , f eljko. 2001. On the syntax-phonology interface: Cliti cization and related phenomena.

Amsterdam: Elsevier Science.

Bod kovi e , f elj ko.in press. On the interpretation of multiple questions. Linguistic Variation

Yearbook.

Bod kovi e , f eljko. in preparation. On wh-islands in South Slavic. Ms., University of Connecticut.

44

Bog kovi h , i eljko and Steven Franks. 2000. Across-the-board movement and LF. Syntax 3: 107-128.

.Brody, Michael. 1995. Lexico-Logical Form: A radically minimalist theory. Cambridge: MIT Press.j

avar, Damir, and Chris Wilder. 1999. "Cliti c third" in Croatian. In Clitics in the languages of

Europe, ed. Henk van Riemsdijk, 429-467. Berlin: Mouton de Gruyter.

Cheng, Lisa. 1997. On the typology of wh-questions. New York: Garland Publications.

Cheng, Lisa. 2000. Moving just the feature. In Wh-scope marking, ed. Uli Lutz, Gereon Müller, and

Arnim von Stechow, 1-19. Amsterdam: John Benjamins.

Cheng, Lisa, and Johan Rooryck. 2000. Licensing wh-in-situ. Syntax 3: 1-19.

Chomsky, Noam. 1995. The minimalist program. Cambridge, Mass.: MIT Press.

Chomsky, Noam. 1999. Derivation by phase. MIT Occasional Papers in Linguistics 18. Department

of Linguistics and Philosophy, MIT, Cambridge, Mass.

Chomsky, Noam, and Howard Lasnik. 1993. The theory of principles and parameters. In Syntax: An

international handbook of contemporary research, Vol. 1, ed. Joachim Jacobs, Arnim von

Stechow, Wolfgang Sternefeld, and Theo Vennemann, 506-569. Berlin: Walter de Gruyter.

Citko, Barbara, and Kleanthes Grohmann. 2000. The (non)-uniqueness of multiple wh-fronting:

German=Bulgarian. Presented at Formal Approaches to Slavic Linguistics 9, Indiana University.

Cole, Peter, and Gabriella Hermon. 1995. Is wh-in-situ really in-situ? Evidence from Malay and

Chinese. In Proceedings of the West Coast Conference on Formal Linguistics 13, 189-204.

Stanford Linguistics Association, Stanford University, Stanford, Calif.

Comorovski, Ileana. 1996. Interrogative phrase and the syntax-semantics interface. Dordrecht:

Kluwer Academic Publications.

Dobrovie-Sorin, Carmen. 1994. The syntax of Romanian: Comparative studies in Romance. Berlin:

Mouton de Gruyter.

Fox, Danny. 1994. Relative clauses and resumptive pronouns in Hebrew: An optimality theoretic

approach. Ms., MIT, Cambridge, Mass.

Franks, Steven. 1993.On parallelism in across-the-board dependencies.Linguistic Inquiry 24:509-

529.

Franks, Steven. 1998. Cliti cs in Slavic. Position paper presented at the Comparative Slavic

Morphosyntax Workshop, Spencer, Indiana, June 1998.

45

Franks, Steven, and Tracy King. 2000. A handbook of Slavic cliti cs.Oxford: Oxford University Press.

Fukui, Naoki. 1997. Attract and the A-over-A Principle. UCI Working Papers in Linguistics 3, 51-

67. Department of Linguistics, University of California, Irvine.

Göbel, Edward. 1998. Focus movement in Romanian. University of Massachusetts Occasional

Papers 21, 83-99. GLSA, University of Massachusetts, Amherst.

Golston,Chris.1995.Syntax outranks phonology:Evidence from Ancient Greek.Phonology12:343-

368

Grewendorf, Günther. 1998. Multiple wh-fronting. Linguistic Inquiry 32: 87-122.

Groat, Erich, and John O’Neil . 1996. Spell -out at the LF interface. In Minimal ideas, ed. Werner

Abraham, Samuel David Epstein, Höskuldur Thráinsson, and C. Jan-Wouter Zwart, 113-139.

Amsterdam: John Benjamins.

Hagstrom, Paul. 1998. Decomposing questions. Doctoral dissertation, MIT, Cambridge, Mass.

Hiramatsu, Kazuko. 1997. A production/judgment asymmetry in children’s negative questions. Ms.,

University of Connecticut, Storrs.

Horvath, Julia. 1986. Focus in the theory of grammar and the syntax of Hungarian.Dordrecht: Foris.

Izvorski, Roumyana. 1993. On wh-movement and focus movement in Bulgarian. Paper presented

at CONSOLE 2, University of Tübingen, December 1993.

Izvorski, Roumyana. 1996. The syntax and semantics of correlative proforms. In Proceedings of

NELS 26, 133-147. GLSA, University of Massachusetts, Amherst.

Izvorski, Roumyana. 1998. Non-indicative wh-complements of possessive and existential predicates.

In Proceedings of NELS 28, 159-173. GLSA, University of Massachusetts, Amherst.

Kayne, Richard. 1994. The antisymmetry of syntax. Cambridge, Mass.: MIT Press.

Kidwai, Ayesha. 1999. Word order and focus positions in Universal Grammar. Ms., Jawaharlal

Nehru University.

Kim, Jeong-Seok. 1997. Syntactic focus movement and ellipsis: A minimalist approach. Doctoral

dissertation, University of Connecticut, Storrs.

King, Tracy Holloway. 1993. Configuring topic and focus in Russian. Doctoral dissertation,

Stanford University, Stanford, Calif.

Kiss, Katalin. 1987. Configurationality in Hungarian. Dordrecht: Reidel.

46

Kiss, Katalin. 1995. Discourse configurational languages. New York: Oxford University Press.

Koizumi, Masatoshi. 1999. Phrase structure in minimalist syntax. Tokyo: Hituzi Syobo.

Kornfilt, Jaklin 1986. The stuttering prohibition and morpheme deletion in Turkish. In Proceedings

of the Turkish Linguistic Conference, 9-10 August 1984, 59-83. Istanbul: Bok aziçi University.

Lambova, Mariana. in press. On A’-movements in Bulgarian and their interaction. Linguistic

Review.

Lasnik, H. 2000. Derivation and representation in modern transformational syntax. In The handbook

of contemporary syntactic theory, ed. Mark Baltin and Chris Colli ns, 62-88. Oxford: Blackwell .

Merchant, Jason. 1999. The syntax of silence: Sluicing, islands, and identity in elli psis. Doctoral

dissertation, University of California Santa Cruz, Santa Cruz, Calif.

Müller, Gereon, and Wolfgang Sternefeld. 1996. A'-chain formation and economy of derivation.

Linguistic Inquiry 27: 480-511.

Napoli, Dona Jo. 1976. The two si’s of Italian. An analysis of reflexive, inchoative, and indefinite

subject sentences in modern standard Italian. Doctoral dissertation, Harvard University.

Nunes, Jairo. 1999. Linearization of chains and phonetic realization of chain links. In Working

minimalism, ed. Samuel Epstein and Norbert Hornstein, 217-249. Cambridge, Mass.: MIT Press.

Nunes, Jairo. 2001. Sideward movement. Linguistic Inquiry 32: 303-344.

Ochi, Masao. 1998. Move or Attract? In Proceedings of the West Coast Conference on Formal

Linguistics 16, 319-333. Stanford Linguistics Association, Stanford University, Stanford, Cali f.

Perlmutter, David. 1972. Evidence for shadow pronouns in French relativization. In The Chicago

witch hunt: Papers from the relative clause festival, ed. Paul Peranteau, Judith Levi, and Gloria

Phares, 73-105. Chicago Linguistic Society, University of Chicago, Chicago, Ill.

Pesetsky, David. 1987. Wh-in situ: Movement and unselective binding. In The representation of

(in)definiteness,ed. Eric Reuland and Alice ter Meulen, 98-129.Cambridge, Mass.: MIT Press.

Pesetsky, David. 1989. The Earliness principle. Ms., MIT, Cambridge, Mass.

Pesetsky, David. 1997. Optimality theory and syntax: Movement and pronunciation. In Optimality

theory: An overview, ed. Diana Archangeli and Terence Langendoen,134-170.Oxford:

Blackwell.

Pesetsky, David. 1998. Some optimality principles of sentence pronunciation. In Is the Best Good

47

Enough, ed. Pilar Barbosa, Danny Fox, Paul Hagstrom, Martha McGinnis, and David Pesetsky,

337-383. Cambridge, Mass.: MIT Press and MIT WPL.

Pesetsky, David. 2000. Phrasal movement and its kin. Cambridge, Mass.: MIT Press.

Pollock, Jean-Yves, Nicola Munaro, and Cecili a Poletto. 1998. Eppur si muove! On comparing

French, Portuguese, and Bellunese wh-movement. Ms., CNRS, Lyon and University of Padua.

Pope, Emily. 1976. Questions and answers in English. The Hague: Mouton.

Reinhart, Tanya. 1997. Interface economy: Focus and markedness. In The role of economy principles

in linguistic theory, ed. Chris Wilder, Hans-Martin Gärtner, and Manfred Bierwisch, 146-169.

Berlin: Akademie Verlag.

Rice, Curt, and Peter Svenonius. 1998. Prosodic V2 in Northern Norwegian. Paper presented at the

West Coast Conference on Formal Linguistics 17, University of British Columbia, Vancouver.

Richards, Norvin. 1997. What moves where when in which language? Doctoral dissertation, MIT,

Cambridge, Mass.

Richards, Norvin. 1998. Shortest moves to (anti-)superiority. In Proceedings of the West Coast

Conference on Formal Linguistics 16, 335-349. Stanford University, Stanford, Calif.

Roberts, Ian. 1997. Restructuring, head movement, and locality. Linguistic Inquiry 28: 423-460.

Rochemont, Michael. 1986. Focus in generative grammar. Philadelphia: John Benjamins.

Rudin, Catherine. 1988. On multiple questions and multiple wh-fronting. Natural Language and

Linguistic Theory 6: 445-501.

Rudin, Catherine. 1993. On focus position and focus marking in Bulgarian questions. In Proceedings

of FLSM 4, 252-265. Department of Linguistics, University of Iowa, Iowa City.

Runner, Jeffrey. 1998. Noun phrase licensing and interpretation. New York: Garland Publications.

Schütze, Carson. 1994. Serbo-Croatian second position cliti c placement and the phonology-syntax

interface. In MIT Working Papers in Linguistics 21, 373-473, MIT, Cambridge, Mass.

Shlonsky, Ur. 1991. Resumptive pronouns as a last resort. Linguistic Inquiry 23: 443-468.

Sinicyn, Misha. 1982. Strange questions in Slavic languages. GLOW Newsletter 8. Dordrecht: Foris.

Stepanov, Arthur. 1998. On wh-fronting in Russian. In Proceedings of NELS 28, 453-467. GLSA,

University of Massachusetts, Amherst.

Stjepanovil , Sandra. 1998. Short-distance movement of wh-phrases in Serbo-Croatian matrix

48

clauses. Presented at the Comparative Slavic Morphosyntax Workshop, Spencer, Indiana, June

1998.

Stjepanovim , Sandra. 1999a. Multiple sluicing and superiority in Serbo-Croatian. In Proceedings of

NELS 29, 145-159. GLSA, University of Massachusetts, Amherst.

. Stjepanovim , Sandra. 1999b. What do second position cliti cization, scrambling, and multiple wh-

fronting have in common. Doctoral dissertation, University of Connecticut, Storrs.

Strahov, Natalya. 2000. A scrambling analysis of Russian wh-questions. Paper presented at Formal

Approaches to Slavic Linguistics 9, Indiana University, Bloomington, February 2000.

Tomi m , Olga Mi n eska. 1996. The Balkan Slavic clausal cliti cs. Natural Language and Linguistic

Theory 14: 811-872.

Wachowicz, Krystyna A. 1974. Against the universality of a single wh-question movement.

Foundations of Language 11: 155-166.

Watanabe, Akira. 1992. Wh-in-situ, Subjacency, and chain formation. MIT Occasional Papers in

Linguistics 2. Department of Linguistics and Philosophy, MIT, Cambridge, Mass.

Department of Linguistics, U-145

University of Connecticut

Storrs, CT 06269

boskovic@sp.uconn.edu